Background
Gastric cancer (GC), which is known for its high heterogeneity and treatment complexity, ranks fifth in terms of morbidity and third in terms of mortality worldwide among all malignant tumors [
1]. In recent years, immunotherapy using immune checkpoint inhibitors (ICIs), mainly against programmed cell death receptor-1 (PD-1) or its ligand, PD-L1, has been one of the biggest advances in the treatment of GC. In particular, in the first-line treatment of metastatic HER2-negative GC, pivotal phase III trials, such as CheckMate 649, KEYNOTE-859, and ORIENT-16, reported that the combination of ICIs with chemotherapy significantly improved overall survival (OS) and progression-free survival (PFS) compared with chemotherapy alone [
2]. However, in both the CheckMate 649 [
3] and ORIENT-16 trials [
4], the survival benefit of immunotherapy seems to come mainly from patients with PD-L1 combined positive score (CPS) ≥ 5, whose proportions (60 and 61%, respectively) were substantially higher than other reports, and such benefit is still controversial in patients with CPS < 5 [
5]. In addition, immunotherapy in second/third-line therapies of GC still has a very limited efficacy (approximately 10% for monotherapy response), regardless of PD-L1 expression [
6,
7]. Novel biomarkers can help in further identifying patients who may benefit from ICI treatment, particularly in the second-/third-line setting.
According to molecular characterization by The Cancer Genome Atlas (TCGA), GC can be classified into four molecular subtypes: chromosomal instability (CIN), Epstein-Barr virus-positive (EBV +), genomically stable (GS), and microsatellite instability (MSI) GC [
8]. EBV + GC accounts for 2–20% of all GC cases and is characterized by an EBV infection, which is usually detected using EBV-encoded small RNA in situ hybridization (EBER-ISH) [
9]. Clinically, EBV testing is often performed based on an undifferentiated phenotype observed by pathologists, described as lymphoepithelioma-like or medullary, and characterized by a dense infiltrate of lymphocytes [
10]. Compared to EBV- GC, EBV + GC has an immune-active tumor microenvironment [
11]. Recently, a phase II trial reported that EBV + GC dramatically responded to second-line immunotherapy with pembrolizumab, with an overall response rate (ORR) of 100% (six patients) [
12]. In contrast, another phase II trial which also enrolled six patients with EBV + GC observed no response to salvage treatment with camrelizumab [
13]. These results indicate that EBV + GC remains highly heterogeneous. In a retrospective study, PD-L1 expression further stratified the outcomes of patients with EBV + GC treated with ICIs [
14]. Such heterogeneity is also reflected by the genomic alterations underlying EBV infection [
15].
In this study, we screened hub genes associated with EBV + GC to identify novel biomarkers of EBV infection and immunotherapy efficacy. We found that CHAF1A, a histone chaperone, was upregulated upon EBV infection. The combination of CHAF1A and current immunotherapeutic biomarkers has the potential to improve clinical practice.
Methods
GC patients with EBV infection data
GC patients diagnosed between January 2020 and August 2023 at the Affiliated Hospital of Jiangsu University (AHJU) were screened for information regarding EBV infection through the following eligibility criteria: gastrectomy, EBER-ISH detection, sufficient tissue for immunohistochemistry (IHC), pathological diagnosis of gastric adenocarcinoma, and no prior history of anticancer therapy (including neoadjuvant therapy). The American Joint Committee on Cancer criteria were used for the clinical and clinicopathological classification and staging. Approval was obtained from the ethics committee of AHJU prior to the study.
Three other GC cohorts with EBV infection data were also used, including those from TCGA [
8], Asian Cancer Research Group (ACRG) [
16], and NCT#02589496 phase II trial [
12]. TCGA and ACRG cohorts were used to screen EBV-associated hub genes and validate their prognostic roles. The AHJU and NCT#02589496 cohorts were used to confirm the association between the hub genes and EBV infection.
Immunotherapy patients
Two cohorts were used to investigate the role of the target gene in the prediction of immunotherapy outcomes. The NCT#02589496 GC cohort enrolled patients to receive second/third-line treatment with pembrolizumab [
12]. The IMvigor210 cohort [
17] included patients with metastatic urothelial cancer (mUC) to receive second-line atezolizumab therapy.
Multi-omic data
Transcriptome data from 34 patients in an additional AHJU GC cohort were used to explore the signaling network associated with the object gene [
18,
19]. Transcriptome data were stored in the European Genome-Phenome Archive (
https://ega-archive.org/), with the identification number EGAD00001004164. Data from other cohorts, including mRNA expression, EBV infection status, tumor mutation burden (TMB), tumor neoantigen burden (TNB), MSI, microsatellite stability (MSS), PD-L1 CPS, and clinical data, have been previously published and were acquired and preprocessed as described elsewhere [
19,
20].
Screening of hub genes associated with EBV + GC
The differentially expressed genes (DEGs) were determined between EBV + GC and EBV- GC using the
limma R package in TCGA cohort, with log
2(fold change) > 0.5 and
p < 0.0001. The prognostic role of the DEGs was evaluated using univariate Cox proportional hazards models, and hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated. DEGs with a significant prognostic impact (
p < 0.001) that were consistent between ACRG and TCGA cohorts were selected. Next, we evaluated the degree of association between DEGs based on semantic similarities in their molecular functions in Gene Ontology (GO) and ranked DEGs based on the average functional similarities between the gene and its interaction partners [
21]. The higher the average functional similarity, the more genes associated with it and the more significant the tested gene.
Cell lines
The human GC cell line HGC-27 and the EBV virus-transformed monkey lymphocyte line B95-8 were purchased from the Type Culture Collection of the Chinese Academy of Science (Shanghai, China).
Generation of EBV + GC cells
B95-8 cells were centrifuged, precipitated, and resuspended in a fresh culture medium. When HGC-27 cells grew to 50% of the culture vessel, B95-8 cells were added and gradually layered on the adherent HGC-27 cell layer from the suspension so that the two cell types began to contact and co-culture. After incubation for 24 h, anti-IgM antibodies and fresh rabbit serum were added to remove B95-8 cells via the immune toxicity response activated by the complement system.
EBER-ISH
ISH was performed using an EBER kit (Zhongshan Jinqiao Biotechnology Co., Ltd.) with an EBER probe according to the manufacturer’s instructions. Briefly, cells were inoculated into a chamber culture slide, fixed with formalin, dehydrated with ethanol after 24 h of culture, and incubated overnight with an EBER probe labeled with digoxin. Diaminobenzidine was used for visualization.
Western blot (WB) and RT-PCR
WB was performed using an anti-CHAF1A (ab126625, Abcam, UK) antibody according to standard protocols. Briefly, after extraction and quantification, total proteins were separated by SDS-PAGE and subsequently transferred onto PVDF membranes (Millipore, Bedford, MA, USA). Then, the membranes were blocked with 5% nonfat dry milk and incubated with ab126625 overnight at 4 °C. Finally, immunoblots were probed with ECL detection reagent (Millipore).
RT-PCR analysis of cDNA was performed using GoTaq qPCR Master Mix and an ABI7300 instrument (Applied Biosystems, USA) according to the manufacturer’s instructions. Briefly, TRIzol (Invitrogen, USA) was used to prepare total RNA, and the Access Reverse Transcriptase-PCR System (Promega, USA) was used to synthesize cDNA.
IHC and multiple-immunofluorescence (mIF) staining
IHC was performed using an anti-CHAF1A antibody (ab126625), with a 2-step protocol. Specialized pathologists calculated the number of positively stained cells and the staining intensity to create grade categories under a microscope. A previously reported semi-quantitative method was used to assess IHC scores [
22]. mIF staining was conducted using the PANO 7-plex IHC kit (Panovue, Beijing, China), section images were reconstructed using the Mantra System (PerkinElmer, Waltham, MA, USA), and quantification of cells in the images was performed using the inForm image software (PerkinElmer). Anti-CD8 (CST70306; Cell Signaling Technology, USA), anti-CD56 (CST3576), anti-CD68 (BX50031; Biolynx, China), anti-HLA-DR (ab92511), anti-panCK (CST4545), and anti-S100 (ab52642) antibodies were used for staining.
Combined score (CS)
In the survival analysis, the optimal cutoff value to define high and low subgroups of TMB or CHAF1A expression with the most significant survival difference was determined using the Survminer R package. The TMB value and CHAF1A expression level were converted to either 1 (high) or 0 (low). EBV infection, MSI, and PD-L1 CPS with a cutoff value of 1 or 5 (CPS1 or CPS5) were converted to either 1 (yes/high) or 0 (no/low). CS was defined as the sum of CHAF1A expression levels with TMB, EBV, MSI, CPS1, and CPS5, ranging from 0 to 2. For response prediction, the receiver operating characteristic (ROC) curve and area under the ROC curve (AUC) were used to evaluate the predictive power of TMB and CHAF1A expression, which was subsequently dichotomized into 1 (high) or 0 (low) based on the optimal threshold of the maximum ROC curve values. Finally, a model for response prediction including all biomarkers was constructed using binary logistic regression with the entry method.
Statistical analyses
According to the need for comparisons between groups, χ2 test, Fisher’s exact probability test, Student’s t-test, and Mann–Whitney U test were adopted. The predictive power of CHAF1A mRNA expression for EBV infection was evaluated by ROC and AUC based on the pROC R package. HRs with 95% CIs were calculated to analyze the independent prognostic value of CHAF1A mRNA expression using multivariate Cox proportional hazard models. The Kaplan–Meier method with the log-rank test was used for survival analysis. Statistical significance was set at p < 0.05. SPSS (version 19.0, Chicago, IL, USA) and R (version 3.6.1) were used for all analyses.
Discussion
Although regimens containing ICIs has become the first-line treatment of GC, there is still no evidence to suggest that patients with GC can benefit from second-line immunotherapy, and the ORR of third-line immunotherapy for GC is extremely low. Recently, owing to unconfirmed clinical benefits, pembrolizumab has been withdrawn as a third-line treatment for GC [
25]. Some biomarkers such as EBV infection, MSI, TMB, and PD-L1 expression have been found to predict immunotherapy efficacy, while studies have reported inconsistent results [
4]. There is still a need to further improve the prediction of immunotherapy response, especially in the second/third-line treatment of GC.
In this study, we identified an EBV-associated gene, CHAF1A, which is upregulated by EBV infection; both its mRNA and protein expression predicted EBV infection in GC. Moreover, CHAF1A alone could predict the prognosis of patients with GC well, but its combination with classic biomarkers, including MSI and TMB, further improved prognostic stratification. Importantly, CHAF1A was a response predictor of immunotherapy for GC, and CS of CHAF1A with EBV, MSI, TMB, or PD-L1 expression further stratified the ORR, which increased with an increase in CS. When all these biomarkers were available, a corresponding model could perfectly predict the response, with an AUC of 0.994. These results indicated that CHAF1A may be a novel immunotherapy biomarker.
CHAF1A is a subunit of chromatin assembly factor-1 (CAF-1), an H3-H4 histone chaperone [
26]. In addition to its epigenetic role, functional versatility of CHAF1A has been reported in GC. The CHAF1A/TCF4 complex directly binds to the promoter regions of c-MYC and CCND1 to enhance their transcriptional activation, thereby promoting gastric carcinogenesis [
27]. Interestingly, HP infection in GC upregulates CHAF1A expression, which is dependent on the binding of specific protein 1 to the CHAF1A promoter [
27]. Recently, CHAF1A is reported to play a role in the infection of human immunodeficiency virus 1 (HIV-1) and be critical in the establishment and maintenance of HIV-1 latency [
28,
29]. In our study, we revealed that EBV infection induced CHAF1A expression, and GSEA suggested that CHAF1A was associated with many infection signaling pathways involving both bacteria and viruses. In particular, the genes involved in the viral carcinogenesis pathway were significantly enriched in the high CHAF1A expression group. Together, these results indicated that CHAF1A participates in pathogen infection and mediates the oncogenic roles of some pathogens.
The role of CHAF1A in anti-cancer immunity remains unclear. Recently, regulators similar to CHAF1A in chromatin organization and remodeling have been reported to play critical roles in anticancer immunity, and have therefore become promising targets for cancer treatment [
30,
31]. Our GSEA showed that CHAF1A was associated with many DNA repair and metabolic pathways. Defective DNA repair increases genomic mutations and instability, which may promote the production of tumor neoantigens and subsequently increase the immunogenicity of tumor cells [
32]. It is also well known that abnormal metabolism in cells of the tumor microenvironment driven by metabolic reprogramming is closely linked to anticancer immunity [
33]. These findings, together with our results showing positive correlations between CHAF1A and MSI, TMB, and immune cell infiltration, suggest that CHAF1A activates anticancer immunity.
Recently, chromatin regulators are revealed to significantly impact tumor response to immunotherapy. The SWItch/sucrose non-fermentable (SWI/SNF) chromatin remodeling complex plays a central role in the coordination of T cell activation and exhaustion [
34]. Inhibition of SWI/SNF results in improved antitumor control, both alone and in combination with immunotherapy [
35]. Genomic alterations in SWI/SNF also affect the response to immunotherapy, and are therefore promising predictive biomarkers [
36]. In our study, the expression of CHAF1A showed the potential to predict immunotherapy response. Similar biomarkers have been widely reported in recent years. However, few of these have been verified in prospective studies, and inconsistent results are concerning. Classic biomarkers such as MSI, TMB, and PD-L1 remain the main basis for clinical decisions. Importantly, CHAF1A was found to be a favorable assistant for the classic biomarkers. The CS of CHAF1A expression with classic biomarkers improved the stratification of both prognosis and immunotherapy outcomes, indicating the possibility of optimizing the use of current biomarkers.
Our study has several limitations. First, the mechanisms by which EBV upregulates CHAF1A expression and the subsequent biological effects of CHAF1A overexpression after EBV infection remain unknown. Second, the mechanisms by which CHAF1A regulates anticancer immunity and determines immunotherapy outcomes remain unclear. Moreover, only one GC cohort undergoing immunotherapy was available for this study, and more such cohorts are required to validate our findings. Finally, prospective validations of a biomarker is necessary.
In conclusion, CHAF1A, a novel biomarker associated with EBV infection, was revealed to be a predictor for prognosis and immunotherapy response in GC. Particularly, CHAF1A had been shown to optimize clinical practice based on current biomarkers by improving their effects. Further validation and research on detail mechanisms are required.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.